US 7995814 B2
The methods described herein are methods to ascertain motion contrast within optical coherence tomography data based upon phase variance. The phase variance contrast observes the nanometer scale motion of scatterers associated with Brownian motion and other non-flow motion. The inventive method of calculating motion contrast from the phase variance can differentiate regions of different mobility based on the motion contrast differences, and can use the phase information to characterize mobility properties of the scatterers. In flow regions, the inventive method for acquiring and analyzing motion contrast can identify the regions as well as characterize the motion. Furthermore, the inventive method can determine quantitative flow estimation, the index of refraction variations, and absorption variations within flow regions.
1. A method to ascertain motion contrast in a sample in an optical coherence tomography system comprising;
acquiring data using the optical coherence tomography system by performing multiple B-scans of the sample, wherein each of said B-scans comprises data acquired during multiple scans over a range of transverse locations;
ascertaining phase variance of the data, wherein said phase variance is ascertained based on data of said B-scans; and
ascertaining the motion contrast in the sample based upon the phase variance.
2. The method of
using the temporal evolution of the measured phase variance of the motion in the sample to identify and characterize mobile scatterers within the images obtained from one or more scans of the sample.
3. The method of
ascertaining the temporal fluctuations in intensity of the data acquired using the optical coherence tomography system and ascertaining the motion contrast based upon the temporal fluctuation.
4. The method of
estimating refractive index variations within flow regions and ascertaining the motion contrast based upon the estimate.
5. The method of
6. The method of
7. The method of
8. The method of
9. A method of ascertaining regions of different mobility in a sample comprising:
ascertaining motion contrast according to the method of
identifying one or more regions of mobility within said sample.
10. The method of
11. The method of
12. A method of diagnosing and treating disease in a patient in need thereof comprising:
ascertaining motion contrast according to the method of
determining blood flow in the area based upon the motion contrast, and
diagnosing the patient based upon the blood flow.
13. The method of
estimating transverse flow velocity based upon the motion contrast.
14. The method of
15. The method of
16. The method of
wherein σ2 Δφ denotes phase variance; Δx denotes transverse motion of the scatterer between phase measurements; vx denotes transverse velocity during the time separation T; and d denotes beam width.
17. A computer readable medium having computer executable instructions for ascertaining motion contrast in a sample in an optical coherence tomography system comprising:
acquiring data using the optical coherence tomography system by performing multiple B-scans of the sample, wherein one of said one or more scans comprises data acquired during multiple scans over a range of transverse locations;
ascertaining phase variance of the data, wherein said phase variance is ascertained based on data of said B-scans; and
ascertaining the motion contrast in the sample based upon the phase variance.
18. An optical coherence tomography system comprising:
a computer readable medium having computer executable instructions for ascertaining motion contrast in a sample comprising:
acquiring data using the optical coherence tomography system by performing multiple B-scans of the sample, wherein one of said one or more scans comprises data acquired during multiple scans over a range of transverse locations; ascertaining phase variance of the data, wherein said phase variance is ascertained based on data of said B-scans; and ascertaining the motion contrast in the sample based upon the phase variance.
19. A method to ascertain motion contrast in a sample in an optical coherence tomography system comprising;
acquiring data using the optical coherence tomography system by performing one or more scans of the sample;
ascertaining phase variance of the data, wherein said phase variance is ascertained independent of intensity data; and
ascertaining the motion contrast in the sample based upon the phase variance.
20. The method of
This application claims the benefit under 35 U.S.C. §119(e) of U.S. Provisional Application Nos. 60/816,431, filed on Jun. 26, 2006, and 60/853,684, filed on Oct. 23, 2006, each of which is incorporated herein by reference.
Optical coherence tomography (OCT) is an optical imaging technique that allows for three dimensional visualization and analysis of structures in a variety of biological systems that are difficult to examine and analyze with other imaging techniques. The embodiments of this invention describe methods to produce motion contrast for a variety of motion types in an OCT system by acquiring and analyzing data according to the inventive method based on phase variance and/or intensity fluctuation analysis.
OCT is a non-invasive optical imaging technique which produces depth-resolved reflectance imaging of samples through the use of a low coherence interferometer system. OCT imaging allows for three-dimensional visualization of structures in a variety of biological systems not easily accessible through other imaging techniques, including but not limited to the retina of the eye.
Vascular visualization and quantitative information of blood flow is very important for the diagnosis and treatment of many diseases. In OCT imaging, a type of phase sensitive analysis called Doppler OCT is the primary form of vascular visualization and diagnostic. Phase is a type of high resolution position measurement of a reflection along the optical path length of the imaging system, which is cyclic of the frequency of half the wavelength of the imaging light. A depth position change of half the imaging wavelength will produce the same phase measurement. Changes in phase are proportional to the axial flow, the flow component parallel to the imaging direction designated by v(cos θ), where v is the velocity of the flow and θ is the angle between the flow direction and the imaging light. Phase noise in the system based on the local signal to noise ratio determines the minimum axial flow measurement, which limits the visualization of flow in cases where v or cos θ is very small. For cases such as in the retina, a lot of flow is nearly perpendicular to the imaging direction such that θ90° and cos θ0. In these cases, the velocity v of the flow must be extremely high to be able to visualize the flow with this method.
The progress of development in OCT is towards faster imaging technologies and techniques in order to image larger regions in the same amount of time. In order to maintain fast imaging speeds, Doppler OCT imaging techniques only use a few successive depth reflectivity measurements called A-scans (the typical number is around 5), and average the phase change between each of them. The limited statistics and the short time between the A-scans (and also the phase measurements) severely limit the minimum observable axial flow, allowing for only visualization of the fastest flows.
Demonstrated variance calculations of the phase changes for this situation do not add additional motion contrast to the images. This lack of additional contrast is because the phase error due to the local signal to noise ratio dominates the calculations in all regions except in the same fast flow regions visualized with the Doppler OCT technique.
Speckle analysis looks at intensity variations of images, which has limited work demonstrated in the field of OCT. Most of the work with speckle in OCT has been directed towards the reduction of speckle artifacts from multiple reflections within the sample to improve image quality. Demonstrated speckle analysis techniques utilize spatial variations of intensity from a single static image to characterize regions and identify regions of flow. These techniques are only capable of analyzing regions much larger than the spatial resolution of the imaging system and have typically been used in non-OCT imaging situations which do not have the depth discrimination capabilities of OCT.
Accordingly, there is a need in the field of OCT for an accurate and efficient method to ascertain flow of biological fluids to assist in the diagnosis and treatment of many diseases. In particular, there is a need to develop a method capable of estimating transverse flow velocity and to ascertain motion contrast with OCT systems.
This application describes the acquisition and analysis methods to produce motion contrast based on phase variance and intensity fluctuation and/or speckle information in an OCT system for a variety of motion types. The phase variance contrast utilizes the temporal evolution of the measured phase variance of the motion to identify and characterize mobile scatterers within the OCT sample images. Acquisition methods are presented which demonstrate a highly efficient acquisition capable of screening for regions of mobility as well as an acquisition capable of quantitative diagnostics of the scatterers.
In another embodiment, the method further comprises using the temporal fluctuations in intensity of the OCT images. The temporal fluctuations in intensity of the OCT images can also be used as another form of contrast to observe the fluctuations associated with flow and absorption changes within the depths of the sample.
The methods and techniques developed herein demonstrate motion contrast within optical coherence tomography images. The motion contrast, and in particular the phase variance contrast is able to observe the nanometer scale motion of scatterers associated with Brownian motion and other non-flow motion. The contrast calculated from the phase variance can differentiate regions of different mobility based on the differences in motion contrast and can use the phase information to characterize mobility properties of the scatterers. For flow regions, the analytical methods described can identify the regions as well as characterize the motion. Quantitative flow estimation can be determined for flow, independent of the orientation relative to the imaging direction. The shadowing of contrast in the phase variance and the intensity fluctuation calculations can be used to determine the index of refraction variations as well as absorption variations within flow regions.
The acquisition methods described herein demonstrate a low efficiency, highly informative diagnostic acquisition as well as an efficient, three-dimensional screening acquisition. The highly efficient acquisition allows for three-dimensional visualization of mobile regions such as a vascular region within the sample. The flexibility of these inventive methods allow for the identification of regions of intermittent flow, such as identification of blood cell motion in the microvasculature. The ability to identify regions of intermittent blood flow can assist in the diagnosis and treatment of a patient in need thereof.
A computer readable medium having computer executable instructions for ascertaining motion contrast in a sample is also contemplated herein. The computer readable medium having computer executable instructions for ascertaining motion contrast comprises acquiring data using an OCT system by performing one or more scan of the sample, ascertaining phase variance of the data, and ascertaining motion contrast in the sample based upon the phase variance. The computer readable medium having computer executable instructions for ascertaining motion contrast may further utilize intensity fluctuation and/or speckle information for ascertaining motion contrast in a sample. The phase variance contrast may utilize the temporal evolution of the measured phase variance of the motion to identify and characterize mobile scatterers within the OCT sample images. The computer readable medium having computer executable instructions for ascertaining motion contrast may also include acquisition methods capable of screening for regions of mobility and quantitative diagnostics of the scatterers.
An OCT machine comprising computer readable media having computer executable instructions for ascertaining motion contrast in a sample is also contemplated herein. The computer readable medium having computer executable instructions for ascertaining motion contrast on the OCT machine comprises acquiring data using the OCT system by performing one or more scan of the sample, ascertaining phase variance of the data, and ascertaining motion contrast in the sample based upon the phase variance. The OCT machine comprising computer readable media having computer executable instructions for ascertaining motion contrast may further utilize intensity fluctuation and/or speckle information for ascertaining motion contrast in a sample. The phase variance contrast may utilize the temporal evolution of the measured phase variance of the motion to identify and characterize mobile scatterers within the OCT sample images. The OCT machine comprising computer readable media having computer executable instructions for ascertaining motion contrast may also include acquisition methods capable of screening for regions of mobility and quantitative diagnostics of the scatterers.
The OCT system used for the data presented herein is a spectral domain optical coherence tomography (SDOCT) setup as shown in
The phase change Δφ(zi,T) measured for a given depth zi for a time separation T is a combination of several factors effecting the measurement:
The phase change Δφ(zi,T) contains not only the individual motion of the scatterer at the depth zi which is designated by Δφmotion,scatterer(zi,T) (which is the motion of interest), but it also contains the bulk relative motion between the sample and the system along the imaging (axial) direction Δφmotion,bulk(T). Δφerror,SNR(zi) is the phase error associated with the SNR of the data calculated at the depth zi. Published experimental results have calculated that the accuracy of measured phase changes is determined by the local signal to noise ratio and is of the form:
One of the methods of determining the bulk sample motion is using an additional motion measurement, like an interferometer for example, to determine the motion of the strongest reflection within the sample. With Fourier domain optical coherence tomography systems, which includes spectral domain optical coherence tomography and swept source optical coherence tomography (also referred to as optical frequency domain imaging), all of the depths of the sample are measured at the same time. With the phase change information available for all of the depth reflections, it is possible to extract the bulk motion for removal.
In some scenarios, there is a strong reflector within the sample that can be used as a stationary reflection to be used as a reference for the bulk phase removal. Many cases do not contain a highly reflective stationary reflector to use, so the entire sample depth must be used to calculate the bulk motion. There are several ways to analyze the phase change information from all of the depths to calculate the bulk sample motion. With the large phase noise associated with depths with no reflectance or signal near the noise level, the mean calculated from all of the phase changes can be distorted due to these low signal terms. Thresholding of the phase data analyzed can reduce the effect of these terms. The mode of the calculated phase changes can be used to determine the bulk motion as well, with an accuracy determined by the parameters used in the mode calculation.
A weighted mean calculation allows for the estimation of the bulk motion for a variety of sample cases. The bulk motion in this method is calculated by Δφmotion,bulk(T)=Σ[w(zi)Δφ(zi,T)]/Σ[w(zi)], where the weighting factor w(zi) is dependant on the type of imaging situation. The weighting factor can be determined by the linear OCT intensity I2(zi), which relies on the highest reflections within the sample being stationary. Using the weighting factor of the OCT amplitude I(zi), the calculation is more sensitive to the phase noise of the low signal terms within the measurement. By incorporating a threshold term into the weighting factor, the phase noise effect from the low signal terms can be reduced in all cases. The weighting term can also contain a spatial dependence to deal with specific sample and motion properties. For example, the stationary regions of a sample below a high velocity flow region can appear non-stationary with the phase change measurements which can require a different weighting above and below the flow regions within the sample. The weighting factor can also incorporate weighting based on the shape of the local intensity function of the depth reflectivity measurement. Artifacts created by phase motion calculations of the side lobes and dips of reflectance profiles can cause unwanted discrepancies which can distort the bulk motion estimation.
With the cyclic nature of the phase measurements, phase changes are forced to be limited between −π and +π. Motions larger than this amount (equivalent to a quarter of the wavelength of the imaging light) suffer from phase wrapping and are miscalculated (a phase change of +π+δ is misinterpreted as −π+δ). In phase measurement cases where the bulk motion of the sample is approximately +/−π, the phase error will cause the calculated phase change distribution to appear similar to the data presented in
After the removal of the estimated bulk motion from the calculated phase changes Δφ(zi,T)−Δφmotion,bulk(T), the variance of this quantity can be approximated by the sum of the variance of the individual components composing the phase change:
The motion of interest as the source of contrast for the phase variance analysis is the scatterer motion phase variance term σΔφ,motion
There are several types of motion which contain components that are observable through the motion scatterer variance measurement of the phase motion which include, but are not limited to:
Each of the above mentioned types of motion have a variance of motion which increases with the time separation between position (phase) measurements. In most cases, the SNR-limited phase error associated with each measured reflection is the limiting factor to the minimum scatterer motion that can be observed. Since this phase error is independent of time, waiting longer between phase measurements allows for the variance of the scatterer motion to increase beyond the limits of the phase error. Further increases to the time separation used for the phase changes will continue to increase the measured phase change variance. This will continue until the calculated phase variance reaches a level comparable to a completely random phase signal. Further increases to the motion of the scatterers should not increase the measured phase variance beyond the completely random phase.
One of the effects that occurs with even longer time separations associated with transverse flow is an appearance of motion shadowing below the regions of flow. This is due to the index of refraction changes which occur within regions of transverse flow. The phase measurement in OCT is not simply a change in the position of a given reflector; it is the change in the optical path length to that same reflector. Therefore, a phase change also measures all of the refractive index variations which have occurred during the time separation T over the entire depth until the measured reflection.
To demonstrate the increase of measured variance motion with increased time separation between phase measurements, the case of Brownian motion is shown. 2% agarose wells are created and filled with an Intralipid solution diluted to match the scattering intensity of the agarose. The agarose is a gelatin which is expected to be stationary in comparison to the mobile Intralipid scatterers.
The phase variance is calculated for this image region for different time separations of the phase change. The images presented in
In order to properly image the contrast created by the scatterer motion, the SNR-limited phase noise should be removed. One of the methods to remove it is to use phase variance calculations from different time separations T1 and T2. If we can assume the phase error from other sources is negligible, the calculated phase variance for these time separations is of the form:
Looking at the time evolution of the phase variance change can allow for the characterization of the mobility of the scatterers.
For the phase variance data measured of the Brownian motion (the data which is less than the random phase variance saturation), the calculated motion is the combination of the time insensitive phase error and the Brownian motion which is of the form:
In order to acquire the data required to identify and/or characterize regions of mobility, phase information from large time separations is required. One of the simplest acquisition methods is to wait at each transverse location, acquiring phase information over time, waiting long enough to acquire the required statistics and the temporal information of the phase change evolution. In terms of OCT scan terminology, an A-scan is the term for a single depth reflectivity measurement. Multiple A-scans acquired at the same transverse location over time is referred to as an M-scan. Multiple A-scans acquired over a range of transverse locations is referred to as a B-scan. The process of waiting at each transverse location creating M-scans, and repeating this over a range of transverse location will be called a MB-scan for this application.
The MB-scan is the easiest way to acquire the statistics and temporal evolution information of the phase variance in order to characterize the mobility of the scatterers. This characterization includes quantification of the flow within regions, along the imaging direction as well as transverse to it (described later). Other characterization capabilities with this information include factors such as the diffusion constant, the scatterer density, and statistical flow information. The only limitation to this method is the inefficiency of the acquisition. The time required to visualize a three-dimensional spatial region with the temporal information of this method would be too long for some applications. A faster acquisition method capable of some of the phase variance contrast visualization is required for screening large three-dimensional regions. Instead of waiting at one location over time until the motion is large enough to produce motion variance contrast, data can be acquired across multiple locations before coming back to the original locations to get additional phase information. This acquisition method is described by acquiring multiple B-scans over time for the same transverse region, which will be called a BM-scan for this application.
The BM-scan is a very efficient method of acquiring phase information with a large time separation without sacrificing total acquisition time of the image. There is limited temporal information with this acquisition method, but the phase information which is acquired allows visualization of slow motions not visible with analysis from successive A-scan acquisitions. Quantitative flow analysis with this method is limited in comparison to the MB-scan. A schematic of some examples of these acquisition methods is shown in
Images created using one form of the MB-scan and the BM-scan over the tail of a zebrafish are demonstrated in
The cases demonstrated in
Transverse Flow Estimation Methods
Park et al.  published results for the expected phase error to occur while transverse scanning over uncorrelated sample reflections as a function of the fraction of the beam diameter scanned between successive phase measurements. According to the Park definition, standard deviation of the phase difference is equivalent to the square root of what is termed phase variance herein.
In order to create phase contrast in the shortest amount of acquisition time, the analysis performed in Park was to determine the limitations of phase contrast between successive A-scans while scanning transversely and creating a B-scan. What was not mentioned was the possibility of using this expected phase error for relative transverse motion between the sample illumination and the sample reflections as a quantitative measure of transverse flow.
If several A-scans are performed at the same transverse position separated by a time T, there will be a phase noise term which appears due to relative transverse motion between the sample illumination and the sample reflections. With the illumination at the same transverse position at successive measurements, the noise term comes from transverse flow, particularly from uncorrelated reflections such as those found in blood. Consider the case of a Gaussian illumination beam with a 1/e2 beam width=d at the focus, and using phase measurements taken at the same transverse position in the same separated by T. The variance of the phase changes is determined for a transverse motion of the scatterer Δx between the phase measurements, created by a transverse velocity vx during the time separation T. Defining the fraction of the beam width as Δx/d, the variance of the phase error due to the transverse motion is calculated to be:
Due to the other error terms which occur in the phase measurements, the accuracy of this technique may be limited to the accuracy of the calibrations of the system and the removal of other phase error terms. From the data presented in Park, the dynamic range of quantitative measure of transverse flow with this analysis appears to be limited to motion which is approximately 20%≦Δx/d≦80% in the time period T. SNR-limited phase noise limit the minimum transverse velocity which can be determined with this method. The upper limit was approximated by the saturation limit of a random phase noise signal limited to be between −π and π. This equation only allows for qualitative measures of flow such that vxT/d>0.8. Increased temporal information can improve the dynamic range of this transverse flow measurement.
In the example of retinal imaging, with a time separation of 40 ms and a focused beam diameter of 20 μm, the dynamic range of quantitative transverse flow calculations is approximately 0.1 mm/s to 0.4 mm/s.
Altering the dynamic range of quantitative flow can be achieved by changing the transverse resolution of the imaging system, or by altering the time separation of measurements T. For example, if retinal imaging was changed to have a time separation of 10 ms with a 30 μm beam diameter, the dynamic range of quantitative transverse flow calculations is approximately 0.6 mm/s to 2.4 mm/s. If the statistics for the phase measurement in this case was sufficient, the phase data acquired with 10 ms time separation can be used to calculate phase variance for a time separation of 20 ms as well (every second phase position measured). This would increase the dynamic range of the quantitative flow measurement by a factor of 2, which in this case results in approximately 0.3 mm/s to 2.4 mm/s.
A second method of determining transverse flow uses the fact that the refractive index variations of transverse flow regions create shadowing artifacts below the flow within the image. If the time separation for the onset can be identified, the depth extent of the flow region can be identified, and the average index of refraction change within the region can be identified. With knowledge of the refractive indices of the scatterers within the flow region, transverse flow rates can be identified. This method is likely to be useful for cases where the flow regions contain multiple types of constituents, like plasma and blood cells within blood vessel flow.
Another method of quantitatively determining transverse flow uses a combination of BM-scans and MB-scans over a region of the sample. The BM-scan phase variance contrast is a highly efficient method of identifying three-dimensional regions of mobility within a sample. With the BM-scans, three-dimensional vasculature can be identified and the directionality of the flow relative to the imaging direction can be used. The MB-scan can be used over the identified vessel of interest and selectively analyzing that specific regional flow. The average axial flow determined from the average phase change can determine the flow within the region. With the location identified by the screening method and the increased statistics available from the MB-scan, small axial flow components can be calculated. With the directionality of the vessel known, the flow within the vessel can be geometrically calculated. The temporal evolution of the phase variance can provide additional correlation to the flow calculation through an estimation of the transverse component of the flow.
Transverse Motion Noise Removal
One of the issues associated with the acquisition methods is the additional noise which accompanies the increased time separation. While the scatterer motion is given more time to move between phase measurements, the sample is also given more time to move as well. The axial motion of the sample is removed through bulk motion removal methods described earlier. The bulk transverse motion is not compensated for with the previous method. As derived in literature, relative transverse motion between the sample and the imaging beam creates a phase error depending on the magnitude of fraction of the imaging beam waist which has been moved . The derivation of the phase noise is based upon the assumption of uncorrelated scatterers within the sample, which is not the case for all reflections within layers samples like the retina.
The approach used to deal with the bulk transverse motion is to assume that the entire contrast image contains the same level of additional phase noise at all transverse locations for the case of the BM-scan. Since the BM-scan acquires data at all transverse locations in a short time frame, all of these points should be experiencing the same motion and the same resulting phase noise. Using all of the contrast data points which are not zero after SNR phase noise removal, the statistics of the contrast image data can be used to try and remove the effect of the transverse motion. For the mean and standard deviation of the non-zero contrast data of the image of μ and σ, respectively, the contrast image C(x,z) can be adjusted through many methods including a removal and normalization of the form:
The phase noise estimation used for the numerical removal for the phase variance contrast images is based on the measured phase noise as a function of the reflectivity signal S2. The measured OCT intensity signal I2 is a combination of the reflectivity signal S2 and the noise signal N2. The averaged OCT intensity signal is of the form:
Additional methods which can be used to remove the transverse motion noise involve using additional statistics to remove any contrast image containing transverse motion noise. Using an external motion trackers or software analysis of the OCT intensity and contrast images, BM-scan data containing significant transverse motion can be identified and removed. Repeating the BM-scan acquisition over the same region when less transverse motion is occurring eliminates the requirement of phase removal analysis.
Intermittent Flow Identification
One of the challenges of motion contrast screening is identifying regions that do not contain scatterers all of the time. OCT requires reflections within the sample to create the phase measurements and some situations do not allow for measurements to be taken at any time. The zebrafish and the segmental vessels are a good example of this reflectivity requirement. The segmental vessels are 7-12 microns in diameter for the 3 dpf (days post fertilization) embryo, which branch off of a larger vessel called the dorsal aorta. Confocal imaging of the embryo has shown that the blood cells are only present within specific locations of these vessels for a fraction of the time. If OCT imaging occurs at a transverse location when there are no blood cells present within the vessel, there is not enough reflectivity from within the vessel to produce a phase contrast signal.
This situation is similar to the microvasculature in the retina, which also contains very small vasculature and does not contain motion contrast at all points in time. As with any random phenomenon, multiple visualization opportunities and increased statistics can assist in the visualization of the events. With the efficiency of the BM-scan acquisition, this method can be repeated over regions expecting an intermittent flow situation, such as microvasculature, in order to visualize the motion contrast.
All of the contrast analysis methods described so far have been directed towards the variance of phase changes over time for scatterers. With the same acquisition methods, there is available data of the OCT intensity information over time as well. Many types of sample properties can result in intensity fluctuations within the image: light coupling changes, source power fluctuations, and relative polarization changes in the interferometer all can cause intensity variations over time. Examples of intensity fluctuations caused by sample motion include, but are not limited to:
While the description above refers to particular techniques for performing the inventive method, it should be readily apparent to people of ordinary skill in the art that a number of modifications may be made without departing from the spirit thereof. The presently disclosed embodiments are, therefore, to be considered in all respects as illustrative and not restrictive.